Model-contrastive federated learning entrenched UWB bi-direction localization through dynamic hexagonal grid construction in indoor WSN environment

Abstract

Indoor wireless sensor networks (WSNs) are necessary for the provision of precise localization services for a variety of applications, including the tracking of assets, the monitoring of the environment, and navigation inside an interior space. Because it offers such high accuracy in distance measurements, ultra-wideband technology is an alternative worth considering for use in indoor localization applications. The vast majority of the activities that are now being done do not take into consideration the nodes in the network in a dynamic manner, and anchor nodes are put at random, which results in difficulties with low connection and poor connectivity. In addition, multi-path fading and incorrect identification of line-of-sight (LOS) and non-line-of-sight (NLOS) during localization, which may occur when the signals are scattered, are additional problems that lead to a poor connection in the current attempts. Numerous prior papers focused solely on sparse techniques like Time difference of arrival (TDoA) and Angle of Arrival (AoA), which are inadequate and have an influence on the precision of the localization process. Within the scope of this investigation, a novel strategy for enhancing UWB-based bi-directional localization in indoor WSNs has been presented. This technique makes use of a complementary pairing of Model-Contrastive federated learning (MCFL) and Dynamic Hexagonal Grid Construction (DHGC). For reducing complexity, we propose a 3D constructed map using hybrid of Tweak Capsule Network (TCapsNet) and Graphical Convolutional Network (GCN). Our proposed method demonstrates significant improvements in localization accuracy in contrast to existing techniques.